Segmentation fault: 11 on macOS

[adlerfaulkner]

Version

v16.16.0

Platform

Locally:
macOS Monterey 12.4
Darwin 20.6.0

Reproduced in a github workflow:
macOS Big Sur 11
Darwin 20.?

Subsystem

No response

What steps will reproduce the bug?

I have a node application which uses the Components.js dependency injection framework. As soon as the call is made to instantiate my application (like this) I get a Segmentation fault: 11.

Full trace using the Segfault handler package:

PID 5340 received SIGSEGV for address: 0x18
0   segfault-handler.node               0x0000000106e3c0ba _ZL16segfault_handleriP9__siginfoPv + 298
1   libsystem_platform.dylib            0x00007ff8027e4dfd _sigtramp + 29
2   node                                0x00000001049ab124 _ZN2v88internal7SlotSet6InsertILNS0_10AccessModeE1EEEvm + 60
3   node                                0x0000000104955022 _ZN2v88internal7Isolate38RunHostImportModuleDynamicallyCallbackENS0_6HandleINS0_6ScriptEEENS2_INS0_6ObjectEEENS0_11MaybeHandleIS5_EE + 218
4   node                                0x0000000104c19ecf _ZN2v88internal25Runtime_DynamicImportCallEiPmPNS0_7IsolateE + 314
5   node                                0x0000000104eac5b4 Builtins_CEntry_Return1_DontSaveFPRegs_ArgvInRegister_NoBuiltinExit + 52
6   node                                0x0000000104f3ae8e Builtins_CallRuntimeHandler + 78
7   node                                0x0000000104e3eeca Builtins_InterpreterEntryTrampoline + 202

I am also getting Segmentation fault: 11 when running a script with node ./dist/index.js that does not use Components.js.

I'm not sure what I should do to debug further.

This seems related to: #25424, #44211, or #43205?

How often does it reproduce? Is there a required condition?

Sometimes if I restart my computer it works.

What is the expected behavior?

No segmentation fault and the application doesn't crash

What do you see instead?

The application crashes

Additional information

No response

[adlerfaulkner]

The problem seems to go away when I comment out imports of @comake/skql-js-engine (which I authored), but it's not clear to me why importing that causes a segmentation fault.

[mscdex]

Perhaps "components.js" is monkey patching or messing with module loader internals in a way that's not compatible?

[adlerfaulkner]

Yeah, not sure because I have other projects which use components.js that don't seg fault. What kinds of things would I look up to see if it's "messing with module loader internals"?

[adlerfaulkner]

I was able to reproduce the error in a github workflow test of an application running my module @comake/skql-js-engine. Thus, this error does not have anything to do with Components.js as I thought previously.

Here you can see the test fails on macos-latest (macOS Big Sur 11) but passes on ubuntu-latest:
https://github.com/comake/skql-js-engine/runs/7830015283?check_suite_focus=true

Here you can see the test fails on macos-12 (macOS Monterey 12) but passes on ubuntu-latest, windows-latest, and macos-latest:
https://github.com/comake/skql-js-engine/runs/7830113462?check_suite_focus=true

On a subsequent run of the tests it passes on all OSs (macos-latest, ubuntu-latest, windows-latest, and mac-12).
https://github.com/comake/skql-js-engine/actions/runs/2857591843

On my local machine running macOS Monterey 12.4 the error happens most times that the test is run but not 100% of the time.

The test is run by npm run test:package which executes a bash script validate-package.sh. The script packs my module and installs it into an example application in ./test/deploy and runs a node script. The segmentation fault is produced by the execution of the node script.

[bnoordhuis]

This issue is very likely a duplicate of #43205. I'll go ahead and close it out.